I vibe coded an AI tool that reads competitors' ads the way a neuroscientist would.

Lilu — Wed, 02 Sep 2026 10:54:50 +0000

I work in advertising and have a background in neuroscience and neuromarketing. Throughout my career the one thing always bugged me: brands pick their emotional angle — fear, status, belonging, safety — mostly based on MMM analysis and historical data, but never on the real behavioural and cultural data.

So I built a platform for competitive neuro-audit. You point it at a market, and it tags competitors' live ads by the behavioral mechanism behind them, maps that onto the country's measured cultural profile, and shows which psychological appeals in the category are overheated and which are still wide open. Then it pretests your chosen appeal implicitly, on a real audience from that market, before you spend a cent on production. So the idea is that everybody can find their customer - it's just the matter of chosing a niche that is still not preoccupied.

Working with Uno — what I liked. Uno gave me credits to build this — uno4.dev

I'm not a developer, and that's exactly why the experience mattered. Uno was genuinely easy to use — I could get things done without fighting the tool — and on the rare occasions I got stuck, support responded fast and kept me moving instead of stalling for days.The speed of turning an idea into something clickable — something I perceive as the biggest win. For a solo founder validating a bet, that's the difference between "someday" and "this weekend."

So I built the working prototype on Uno.... but everything broke down when I tried to create a unified database for various markets (MENA, APAC, and elsewhere) with behavior patterns and perception hooks. Actually, it was my own problem, not the platforms.

Many studies were contradictory, and the system constantly labeled the data differently, making it difficult to develop a unified template for each country and market. Comparing clients was also difficult, as the patterns for different products and services also varied, and the model needed to be trained in multiple layers, taking these factors into account.. I'm still figuring it out but I never regret trying to do this!

My point is - don't give up and always date to build something on your own, even if you're not good in coding as I am.

DEV Community: Lilu

I vibe coded an AI tool that reads competitors' ads the way a neuroscientist would.